Face recognition based customer authentication by using deep learning techiques for detecting atm fraud

Today, cash dispensers are widely used for bank customers to perform their financial business. The cash dispensers are put into use after providing customer verification with a magnetic card and a four-digit PIN. This provides an opportunity for fraudsters to steal money using the hardware of cash dispensers. As the number of digital banking platforms increased, bank customers have been able to open accounts without having to physically visit the bank branches. Similarly, bank cards can be delivered to customers by courier without having to visit the branch. This gives evil-minded courier employees the opportunity to steal card information. In addition, bank customers typically set their four-digit PIN code as the date of birth, the license plates of the cities they live in, or their national identification numbers. Fraudsters can guess this code or identify themselves as bank employees and can easily obtain them. Due to such disadvantages in credit and debit cards, fraudsters have the potential to easily steal money from bank customers. In this work, in order to prevent related fraud in ATMs, an authentication method by using face recognition technologies in addition to the standard password based authentication method of banks has been introduced. When the customer opens an account at the bank, the face samples taken from the customer will be passed through a deep learning model and a unique identity will be produced and recorded. Afterwards, when the customer is processing at the cash dispenser, the faces taken again will be passed to the deep model and the result will be compared with the recorded identity. When this comparison result is below the threshold value, customer verification will be approved. Otherwise, access to the cash dispenser will not be possible even with card and PIN information. Thus, it will be ensured that banks provide a much safer platform in the services they provide through ATMs.